-
Notifications
You must be signed in to change notification settings - Fork 33
/
test.py
170 lines (139 loc) · 5.6 KB
/
test.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
import numpy as np
import matplotlib
# Force matplotlib to not use any Xwindows backend.
matplotlib.use('Agg')
import matplotlib.pyplot as plt
import pygco
# def get_uniform_smoothness_pw_single_image(img):
# """
# Generate uniform smoothness pairwise potential for a single image of size
# img. Pixel indices are assumed to be in row-major order.
# In a uniform smoothness pairwse potential, for any pair of neighboring
# pixels i and j, p(i,j) = 1
# img: a tuple of two integers H,W for height and width of the image
# return: edges, edge_weights
# edges is a E*2 matrix, E is the number of edges in the grid graph.
# For 4-connected graphs, E=(H-1)*W + H*(W-1). Each row is a pair of
# pixel indices for an edge
# edge_weights is a E-dimensional vector of 1's.
#
# see: https://github.com/kelvinxu/Segmentation-Code/blob/master/imgtools/pairwise.py
# """
# H, W = img
# E = (H - 1) * W + H * (W - 1)
#
# edges = np.empty((E, 2), dtype=np.int)
# edge_weights = np.ones(E, dtype=np.single)
# idx = 0
#
# # horizontal edges
# for row in range(H):
# edges[idx:idx+W-1,0] = np.arange(W-1) + row * W
# edges[idx:idx+W-1,1] = np.arange(W-1) + row * W + 1
# idx += W-1
#
# # vertical edges
# for col in range(W):
# edges[idx:idx+H-1,0] = np.arange(0, (H-1)*W, W) + col
# edges[idx:idx+H-1,1] = np.arange(W, H*W, W) + col
# idx += H-1
#
# return [edges, edge_weights]
def test_gc():
""" """
gc = pygco.gco()
gc.createGeneralGraph(3, 2, True)
gc.handle is not None
gc.destroyGraph()
def test_integer():
""" """
unary = np.array([[2, 8, 8],
[7, 3, 7],
[8, 8, 2],
[6, 4, 6]])
edges = np.array([[0, 1], [1, 2], [2, 3]])
edge_weight = np.array([3, 10, 1])
smooth = 1 - np.eye(3)
labels = pygco.cut_general_graph(edges, edge_weight, unary, smooth,
n_iter=1)
np.array_equal(labels, np.array([0, 2, 2, 1]))
def test_float():
""" """
unary = np.array([[0.0, 1.0, 2.0],
[4.0, 1.0, 0.0],
[1.0, 0.0, 2.0]])
edges = np.array([[0, 1],
[1, 2],
[0, 2]]).astype(np.int32)
smooth = (1 - np.eye(3)).astype(np.float)
edge_weights = np.array([2.0, 0.0, 0.0])
labels = pygco.cut_general_graph(edges, edge_weights, unary, smooth,
n_iter=-1, algorithm="swap")
np.array_equal(labels, np.array([0, 2, 1]))
def test_grid():
""" """
annot = np.zeros((100, 100))
annot[:, 50:] = 2
annot[25:75, 25:75] = 1
noise = annot + np.random.randn(100, 100)
unary = np.tile(noise[:, :, np.newaxis], [1, 1, 3])
tmp = unary[:, :, 1]
tmp = (tmp - 1)
tmp[annot == 0] *= -1
unary[:, :, 1] = tmp
unary[:, :, 2] = 2 - unary[:, :, 2]
fig = plt.figure(figsize=(unary.shape[-1] * 4, 4))
for i in range(unary.shape[-1]):
plt.subplot(1, unary.shape[-1], i + 1)
plt.imshow(unary[:, :, i], cmap="gray", interpolation="nearest")
fig.tight_layout(), fig.savefig('./images/grid_unary.png')
pairwise = (1 - np.eye(3)) * 10
labels = pygco.cut_grid_graph_simple(unary, pairwise, n_iter=-1)
fig = plt.figure(figsize=(2 * 4, 4))
plt.subplot(1, 2, 1), plt.title('original annotation')
plt.imshow(annot, interpolation="nearest")
plt.subplot(1, 2, 2), plt.title('resulting labeling')
plt.imshow(labels.reshape(100, 100), interpolation="nearest")
fig.tight_layout(), fig.savefig('./images/grid_labels.png')
def test_binary():
""" """
img = np.random.randn(100, 100)
img[25:75, 25:75] += 2
img -= 1
# !!! Be careful when doing this concatenation,
# it seems 'c_' does not create a copy
#u = np.c_[img.flatten().copy(), -img.flatten().copy()]
unary = np.c_[img.reshape(img.size, 1), -img.reshape(img.size, 1)].copy()
fig = plt.figure(figsize=(unary.shape[-1] * 4, 4))
for i in range(unary.shape[-1]):
plt.subplot(1, unary.shape[-1], i + 1)
plt.imshow(unary[:, i].reshape((100, 100)), cmap="gray", interpolation="nearest")
fig.tight_layout(), fig.savefig('./images/binary_unary.png')
# edges, edge_weights = get_uniform_smoothness_pw_single_image(img.shape)
smooth = 1 - np.eye(2)
unary = np.tile(img[:,:,np.newaxis], [1, 1, 2])
unary[:, :, 0] = img
unary[:, :, 1] = -img
unary_new = unary.reshape((100, 100, 2))
assert np.abs(unary - unary_new).max() == 0.
assert not (unary != unary_new).any()
# y = pygco.cut_grid_graph_simple(unary, pw_cost*0, n_iter=-1)
# labels = pygco.cut_grid_graph_simple(unary_new + np.random.
# randn(unary.shape[0], unary.shape[1], unary.shape[2])*0, pw_cost*0, n_iter=-1)
labels = pygco.cut_grid_graph_simple(unary_new + np.zeros(unary_new.shape),
smooth, n_iter=-1)
labels_0 = pygco.cut_grid_graph_simple(unary_new, smooth * 0, n_iter=-1)
fig = plt.figure(figsize=(3 * 4, 4))
plt.subplot(1, 3, 1), plt.title('image')
plt.imshow(img, interpolation="nearest")
plt.subplot(1, 3, 2), plt.title('labeling (smooth=1)')
plt.imshow(labels.reshape(100, 100), interpolation="nearest")
plt.subplot(1, 3, 3), plt.title('labeling (smooth=0)')
plt.imshow(labels_0.reshape(100, 100), interpolation="nearest")
fig.tight_layout(), fig.savefig('./images/binary_labels.png')
# if __name__ == "__main__":
# test_gc()
# test_integer()
# test_float()
# test_grid()
# test_binary()